Vibration Analysis in Production Testing

The dynamic properties of structures can be described with the aid of eigenfrequencies, attenuation and eigenshapes. Suitably measured, analyzed and classified, these properties can quickly determine quality for pass/fail selection of manufactured components. For many production inspection applications, Laser-Doppler vibrometers offer significant advantages over conventional, structure-born noise sensors, such as contact accelerometers. Apart from the high dynamic range, wide bandwidth and small spatial measurement spot, it is the non-contact, nonreactive measurement of a laser vibrometer that plays a decisive role. There are many high performance methods available to analyze vibration signals (see table on page E24). In this tutorial, the basic principles of frequency analysis are introduced including two versions of frequency band analysis: short-term Fourier analysis and order analysis. Polytec GmbH Polytec-Platz 1-7 D-76337 Waldbronn Tel. +49 (0)72 43 6 04-0 Fax +49 (0)72 43 6 99 44

Advancing Measurements by Light

the position of the spectral frequency lines is usually evaluated directly by the frequency analysis. e Ϯj2␲kn / N Figure 1: The increased resolution of a long measurement versus a short measure-ment time. but. operational vibration analysis is not entirely concerned with the structural resonances. In simple applications. the status of the test item is determined. the phase of the individual frequencies nf0 is of no importance. 1. 1). Hanning. red: f⌬ = 10 Hz Figure 3: Digitalization of a 4 kHz signal.. . For real signals. green: sample frequency 10 kHz. broad frequency line (Fig.56 · fC. 3. a factor of 2. green: f⌬ = 1 Hz. highest frequency occurring in the signal fmax.g. 2. The frequency resolution f⌬ = fA / N = 1/ T = 1/ NT⌬ solely depends on the sample frequency fA and the number of data points N and thus on the duration of the measurement. N needs to be calculated for symmetry reasons. T. red: sam-ple frequency 5 kHz Figure 4: Results of frequency analysis of both signals from fig. also.56 is usually selected in the digital signal processing and applied to the cutoff frequency of the selected low pass filter to determine an appropriate sampling frequency. such as testing E 22 . red: pseudo-frequency through Aliasing = e Ϯj2␲kn (k+␣N) / N.1 s measurement duration. From these characteristic frequencies. fA = 2. 3. In material testing. Usually in frequency analysis. the analog signal must be band-limited with a low-pass filter prior to digitalization. Because of the periodicity of the discrete Fourier spectrum in N. 2). Leakage is noticed by additional frequencies in the spectrum. red: 0. With a discrete Fourier transformation. To prevent this. the forced vibrations Figure 2: Results of both signals from fig. Leakage must be expected if the product of the observation period T and the frequencies to be displayed fn is not an integer. only the characteristic resonant frequencies of the component are ascertained. This vibration content depends on both the resonant frequencies of the components and the driven movement velocity that often corresponds to an RPM value. frequency parts outside the cutoff frequency. 4). Another effect in calculating the Fourier spectrum of a discrete signal is the so-called Leakage. . If the measurement time selected is too short.. a signal s(tk ) = s(kT ⌬) with N samples is mapped onto N discrete frequencies fn = nf⌬ (Fig. Discret frequency variables fn = nf⌬ with f⌬ = 1/ T are the only way to display frequencies that are an integer multiple of the frequency resolution without error. Frequency Band Analysis In contrast to material testing. then individual frequencies close to each other will no longer be resolved. Since all real filters have a transitional area between the pass band and the stop band. green: real fre-quency. To prevent aliasing for real signals.Frequency Analysis Automatic digital analysis of vibration signatures for product final inspection is most commonly done using frequency analysis. fC. This can be avoided by selecting the sample frequency fA according to Shannon's sample theory: to be at least twice as big as the Hamming) before the Fourier transformation. of the filter are not fully attenuated and can lead to aliasing artifacts. 1. leading to an effect known as Aliasing (Fig. green: 1 s measurement duration shape or continuously with a shaker. the power density spectrum |S(fn)|2 is analyzed. only the Fourier spectrum S(fn) for n = 0. Independent of whether the components are excited in a pulse generated through the interaction of assembled components. It can be reduced if the discrete time signal is weighted with a suitable window function (e. but will be seen as a single. ␣ ʦ N there can be an overlap of individual spectra.

modulation analysis is based on frequency analysis of the rectified signal. i. in modulation analysis the frequency analysis is usually applied to several carrier frequency bands. Bi. the individual frequency lines are not analyzed. even with a slight deviation in the RPM value. Bi+1 is constant and equal to the proportion qf of the corresponding band average frequencies. qB. in modulation analysis. m = 0. Since carrier frequencies with a different frequency can be modulated differently. If qB = qf = 21/3 applies. M Յ N Figure 5: Green: spectrum of a signal. actuators or attenuators. the average performance is Figure 7: Fourier transformation.. The short-time Fourier transformation provides a constant time-frequency resolution. whereby the average frequencies of the bands are arranged equidistantly. but continuously for joined time periods. the vibration characteristics continually change. the mechanical construction has no stationary condition that can be attained which means that the test sample is in different operating modes at different points in the process. the proportion. instead. on the left: output signal. where the bandwidth of the frequency intervals increases with increasing average band frequency. 5 (frequency axis is llogarithmic) analysis. the signal spectra are divided up into frequency intervals which are seamlessly strung together. are examined. The power density is determined for every frequency band and the so-called narrowband spectrum is used for analysis. the degree of modulation. The partitioning of the frequency bands for one-third octave band analysis is oriented towards the subjective perception of the human ear. fm. In this case.electric motors or cooling fans. of the respective neighboring frequency bands.e. the calculated frequency bands are designated as 1/3 octave bands. to be able to carry out detailed analysis of the vibration characteristics at certain points in time. called frequency bands. E 23 . short-time Fourier analysis is often used (Fig. Identical products with similar behavior are analyzed at a constant RPM and can be reliably qualified. This means that during operation. the resolution of the time period/bandwidth product is equal for all values n and m. not determined for the whole signal. In principal. In doing so. fm. 7). groups of frequencies . the ratio between the carrier frequency amplitude and the modulation amplitude is also determined. 1. such as testing gearboxes.. the discrete time signal is weighted in a window function w(k – m) which depends on a translation parameter and subsequently the Fourier transformation is applied to it. An initial impression of the change in the vibration characteristics at different operating modes allows analysis of the level progression. The window function w(k – m) then ensures that just one time section of the signal s(k) is transformed at a time. How-ever. Therefore. The time period of the signal section observed is defined by the width of the window function and determines the spectral bandwidth.. blue: one-third octave band level of the spectrum Figure 6: One-third octave band level spectrum from fig. . and the analysis as one-third octave band Modulation Analysis is used to evaluate the spectral composition of the envelopes of signals caused by amplitude modulation. i+1. M – 1. In the case of narrow band analysis. Short-time Fourier Analysis For certain applications. Apart from the spectral analysis of the envelopes. In contrast to narrow band analysis. proportional band analysis uses spectra that are divided into joined frequency intervals. on the right: spectrogram (short-time FFT) In short-time Fourier analysis. in which doubling the physical frequency is seen as a proportional frequency increase (Figure 5/6). i .

such as car engines or gearboxes. Polytec’s ruggedized laser-Doppler vibrometers and easy-to-use Quality Control Software offer optimal solutions for more efficient inspection. it is not possible to analyze events which occur briefly using short-time Fourier transforms. frequency analysis does not provide meaningful results. including Fourier transformation and octave analysis.g. solidity testing.g. Due to the permanently changing operating mode. pulse hammer) continuous. single-channel signal analyzer that is designed for pass-fail evaluation of acoustic signatures emanating from a product or machine under test. operational vibration analysis and process control. a pass/fail decision is calculated by comparing the vibration response of each test sample with the limits corresponding to that product test. QuickCheck – Hardware and Software QuickAnalyzer – Efficient Analysis Package for Acoustic Quality Control of Vibration Signals QuickCheck is a multi-channel. flaw (crack) testing.polytec. enhancing process throughput.. not constant subject to changing operating modes run-up or run-down The complete Polytec Tutorial Series can be found as downloadable PDF files on www. non-contact vibrometers. It is designed for research and development of process and quality control applications. Conclusion High-performance. The pass/fail decision is indicated on the display. transferred to the production controller and stored in a log file for future statistical evaluation. which are tested and operated over a long period. Typical applications include noise and vibration testing. Many methods. In E 24 . PC-based software for semi or fully automatic process monitoring and control using the vibration behavior of the product. periodic (e. continuous. are included as standard software for analyzing vibration signals. the user can quickly and efficiently differentiate between unique signal Application Material testing Type of excitation external Excitation pulse-shaped (e. 100% in-line quality inspection. an “audition” module enables digital filtering and audio playback of the saved signal. The software can then group measurement data into fault classes which are clearly arranged and displayed in the analysis window. entered into the software and stored in a database. During the test process. enable in-line 100% quality inspection and control of components and products.polytec. material testing. combined with appropriate signal analysis. The range limits for each product test are assigned a priori during quality examinations of good and bad products. In such cases. QuickAnalyzer is a PC-based. continuous. Such machines are usually tested in one or more cycles with continuous run-up. The software has various modeling techniques that are available for analyzing measurement data. To learn more visit: www. with shakers) Function – crack detection – resonance testing – resonance testing – engines – cooling fans – bearings – drives – actuators – attenuators – combustion engines – gearboxes Analysis – frequency analysis – frequency analysis – total level – narrowband analysis – 1/3 octave band analysis – modulation analysis – level progression – short-time Fourier analysis – order analysis Noise testing Operational vibration analysis self-excitation. constant at a constant velocity self-excitation. and also to the order (multiple of the RPM) in place of the frequency. By simply selecting a new analysis window.Order Tracking For high performance machines. order analysis is used in which the vibration amplitude instead of the short-time Fourier transform is applied to the RPM value in place of the time. lowering product cost and increasing customer perception of product quality.

Sign up to vote on this title
UsefulNot useful